Claude Code v2.1.200: Key Fixes and Permission Mode Changes
Anthropic released Claude Code v2.1.200 with several behavioral changes and bug fixes targeting background agents, permission handling, and terminal compatibility. This release is available now via GitHub.
Permission Mode Changes
- AskUserQuestion dialogs no longer auto-continue by default. Opt into an idle timeout via
/config. - The "default" permission mode is now Manual across CLI,
--help, VS Code, and JetBrains. Use--permission-mode manualor"defaultMode": "manual"in config (note:defaultis still accepted but maps to manual).
Bug Fixes
- Startup crash fixed when
disabledMcpServersorenabledMcpServersin.claude.jsonis set to a non-array value. - Background sessions now properly resume after sleep/wake or reopening a stalled session, and a cancelled turn with Esc no longer re-runs after a stall respawn.
- Background agents no longer fail to start due to a stale
daemon.lockwhose PID the OS reused. - Daemon handover now uses the version's embedded build timestamp to prevent older builds from taking over the daemon.
- Roster issues fixed: transient corruption no longer permanently disables orphan cleanup, older binaries preserve fields written by newer versions, socket auth tokens are no longer stripped during daemon restarts.
- Subagents cut off by a rate limit before producing text now fail cleanly instead of returning empty results.
- Control bytes from background-agent output no longer reach the terminal in agent view.
claude agents --plugin-dir <dir>now works when the flag is placed afteragents.- Project-scoped plugins load correctly from git worktrees of the same repository.
/mcpserver list now tracks focus for screen readers and magnifiers.- Voice dictation no longer shows a misleading "Voice connection failed" message when no audio is captured.
Improvements
- Rendering flicker under tmux 3.4+ fixed by enabling synchronized terminal output.
- Screen-reader output improved: decorative glyphs hidden, transcript symbols read as short labels, nested tables read as
Header: value.lines. - Install script now explains when installation is killed by the system running out of memory.
📖 Read the full source: GitHub Claude-Code
👀 See Also
AI Reimplementation of chardet Library Raises Copyleft Licensing Questions
Dan Blanchard used Anthropic's Claude to reimplement the chardet Python library from scratch, changing the license from LGPL to MIT. The resulting code shows less than 1.3% similarity to previous versions, sparking debate about whether AI-assisted reimplementation erodes copyleft protections.
Taalas' HC1: Accelerating AI Inference with Custom Silicon
Taalas unveils the HC1 platform, providing the world's fastest, low-cost AI inference using model-specific hardware design. Achieves 17K tokens/sec on Llama 3.1 8B.
Two new models appear on OpenRouter, possibly DeepSeek V4 variants
Two new models named healer-alpha and hunter-alpha have appeared on OpenRouter, with specifications matching leaked details about DeepSeek V4. Initial testing shows both models perform well in roleplay scenarios with no message filtering and faster token generation than GLM 5.0.
Claude Code v2.1.91 Updates: Agent Design Patterns, Memory Rules, and Tool Improvements
Claude Code v2.1.91 adds a reference guide for agent design patterns covering tool surface design, context management, and caching strategies. The update simplifies memory selection rules, adds security monitoring for memory poisoning, and improves tool descriptions for Edit, ReadFile, and Write operations.